nano-banana
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using the
reflyCLI and standard utilities likejqandopento interact with workflows and files. - [EXTERNAL_DOWNLOADS]: The skill downloads generated image files from the platform's official domain (
refly.ai) to the user's local directory ($HOME/Desktop). - [INDIRECT_PROMPT_INJECTION]: The skill accepts external text input via the
image_promptfield which is then processed by a downstream image generation workflow. - Ingestion points: User-provided
image_promptandimage_styleinSKILL.md. - Boundary markers: None observed in the implementation steps.
- Capability inventory: The skill has the capability to execute shell commands (
refly,jq,open) and write files to the local file system. - Sanitization: Input is passed through standard JSON structures for the CLI command.
Audit Metadata